Designing Your First Robot with AI: A Kindly IDE Walkthrough

What if you could describe a robot in plain English and get a fully validated design back in minutes? That's exactly what Kindly IDE does. In this walkthrough, we'll take you from a blank canvas to an export-ready robot.

Step 1: Describe What You Need

Open Kindly IDE and type a natural language description into the AI panel. Something like: "A 6-DOF robotic arm with a parallel gripper, mounted on a table. Each joint should have position control. The arm should have a reach of about 800mm."

The AI agent parses your description and generates a complete kinematic chain: links with proper dimensions, joints with realistic limits, inertial properties based on assumed materials, and collision geometries that match the visual meshes.

Step 2: Validate the Physics

Once the structure is generated, the built-in physics linter runs automatically. It checks for common errors that would cause simulation failures:

• Inertia tensors — flags unrealistic or zero-mass links
• Joint limits — catches reversed upper/lower bounds or limits that would cause self-collision
• Collision geometry — warns about overlapping primitives or missing collision meshes
• Kinematic loops — detects unsupported closed chains

Errors appear inline in the viewport with suggested fixes. Click to auto-correct, or adjust manually in the property panel.

Step 3: Configure ros2_control

If you're planning to run this on real hardware, the ros2_control generator creates everything you need: hardware interface plugins with the correct read/write patterns for your protocol (UART, CAN, I2C, or SPI), controller YAML configs for position/velocity/trajectory control, and a launch file that ties it all together.

For simulation-only workflows, you can skip this step entirely—the URDF export already includes the transmission and gazebo tags needed for Gazebo or MuJoCo.

Step 4: Preview in the Viewport

The 3D viewport renders your robot in real time using React Three Fiber with Rapier3D physics. You can:

• Drag joints to test range of motion
• Enable IK mode to check reachability
• Toggle gravity to see if the robot holds its pose
• Run collision detection to verify clearances

Step 5: Export

When you're satisfied, export to URDF (for ROS/ROS2), SDF (for Gazebo), or MJCF (for MuJoCo). The exporter handles mesh embedding, material mapping, and format-specific quirks so the file works out of the box in your target simulator.

You can also export a complete ROS2 package with CMakeLists.txt, package.xml, launch files, and config directories pre-populated—ready for colcon build.